Token导航 LogoToken导航TokenDH.com
研究检索敏感数据clawhub未标认证来源可访问clear审计提醒

pilot-service-agents试点服务 Agent

Agent Skill

pilot-service-agents 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

763

周安装

49

GitHub Stars

公开资料未说明

下载量

384
OpenClaw

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

MIT-0

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:pilot-service-agents(试点服务 Agent)
来源仓库:https://github.com/teoslayer/pilot-service-agents
安装命令:
openclaw skills install pilot-service-agents
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install pilot-service-agents

简介

用于发现并调用约 370 个包装现实世界 API 的数据代理。

  • 适合快速集成地图、天气、金融等第三方服务的场景。
  • 所有代理均通过 Pilot 协议暴露标准化接口。
  • 安装命令:openclaw skills install pilot-service-agents。
  • 部分高级功能可能需要额外付费订阅或 API 密钥授权。

SKILL.md

name
pilot-service-agents
description
>
tags
license
AGPL-3.0
compatibility
>
metadata
author
vulture-labs
version
1.0
openclaw
requires
bins
homepage
https://pilotprotocol.network
allowed-tools

pilot-service-agents

Every service agent on Pilot Protocol exposes the same four-command interface over the overlay, so once you understand one you understand them all. The catalogue itself is served by a directory agent called list-agents — always start there.

The flow: discover → pick → invoke → read inbox

pilotctl send-message list-agents --data "/data {filters}"   (find agents)
pilotctl send-message <hostname>   --data "/help"            (read contract)
pilotctl send-message <hostname>   --data "/data {filters}"  (fetch data)
pilotctl inbox                                               (read response)

The send-message call returns an ACK immediately; the actual response comes back as an inbox message a few seconds later from the agent you called.

Discovering agents via list-agents

list-agents is the directory. Talk to it like any other agent.

List all commands it supports

pilotctl --json send-message list-agents --data "/help"
pilotctl --json inbox

Filter the catalogue

pilotctl --json send-message list-agents --data '/data {"category":"academic","limit":20}'
pilotctl --json inbox

Supported filter fields on list-agents /data:

fieldtypemeaning
categorystringacademic, geo, finance, news, health, science, security, …
tierstringfree (no per-request cost) or premium (paid, higher quality, e.g. Google Cloud)
searchstringsubstring match on hostname + category + description
hostnamestringexact hostname match
limitintmax items (default 50, max 500)

The response envelope contains items, count, total, groups (counts per category), and tiers.free / tiers.premium buckets with their disclaimers.

Gemini summary of matches

pilotctl --json send-message list-agents --data '/summary {"category":"flights"}'
pilotctl --json inbox

Free-text question over the catalogue

pilotctl --json send-message list-agents --data 'which agent gives me Formula 1 race results?'
pilotctl --json inbox

Talking to a specific service agent

Once you have a hostname, the same four commands work on every agent:

messagewhat it does
/helpprints the filter schema + pagination + example call
/data {json}fetch real data, returns a normalised envelope
/summary {json}same fetch, piped through Gemini for prose
<any free text>Gemini picks filters and fetches

Filter shape for /data is agent-specific. Always read /help first — it lists every filter with type, required flag, default, and description. Premium agents (hostname prefix gcp-) announce themselves with a [PREMIUM] tag in /help.

Canonical response envelope

Every /data response has this shape:

{
  "source": "<hostname>",
  "items":  [...],
  "count":  <int>,
  "total":  <int|null>,
  "page":   <int|null>,
  "next":   <cursor|null>,
  "truncated": <bool>,
  "upstream_url": "<resolved upstream URL>"
}

Pagination

If an agent's /help shows Pagination: page|offset|cursor, pass the knob in your filter JSON. The knobs the agent recognises are in the same /help output. Omit pagination and you get the upstream's default page.

Workflow Example

Answer "what aircraft are currently over Manhattan?" with zero upfront API knowledge:

# 1. Find the right category
pilotctl --json send-message list-agents --data '/data {"category":"flights","limit":5}'
sleep 5
pilotctl --json inbox   # pick e.g. adsb-lol-latlon

# 2. Read its filter contract
pilotctl --json send-message adsb-lol-latlon --data '/help'
sleep 5
pilotctl --json inbox

# 3. Query with the filters you just learned
pilotctl --json send-message adsb-lol-latlon \
  --data '/data {"lat":40.78,"lon":-73.97,"radius":8}'
sleep 5
pilotctl --json inbox

What to expect across the catalogue

The catalogue grows; treat every hostname list as a snapshot. To stay current, always start with list-agents /data — never hard-code a list.

  • Free agents — community API wrappers. Best for volume, exploration,

unauthenticated lookups. Rate-limited on the upstream's side.

  • Premium agentsgcp-* hostnames, backed by Google Cloud credits

managed centrally. Higher accuracy, SLA-grade freshness, real cost per call.

  • Every agent caches upstream responses for ~60s and serves stale data on

upstream errors ("cached": true, "cache_age_seconds": N fields appear in the envelope when that happens).

  • Authentication is handled agent-side. You never hold an API key.

Related skills (per category)

There is a dedicated pilot-service-agents-<category> skill for each category in the catalogue — read one of those once you know which kind of data you need. Discover the list with:

pilotctl --json send-message list-agents --data '/data {}'

Then look for the matching pilot-service-agents-* skill.

Dependencies

Requires pilot-protocol core skill, a running pilotctl daemon joined to network 9 (pilotctl --json network join 9), and the list-agents directory agent reachable on the overlay.

适合场景

01

OpenClaw 用户查找和安装 Skill 时

02

用户想查找某类 Agent Skill 时

03

需要根据任务场景推荐可安装能力包时

04

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

补充不同宿主或平台的使用分布数据

能力 5

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

OpenClaw

84.8%
按下载量换算326

安全审计

VirusTotal

可疑

ClawScan

通过

Static analysis

通过

权限和风险

敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills